Blockchain-based Public Auditing with Deep Reinforcement Learning for Cloud Storage
Jiaxing Li,Jigang Wu,Lin Jiang,Jin Li
DOI: https://doi.org/10.1016/j.eswa.2023.122764
IF: 8.5
2024-01-01
Expert Systems with Applications
Abstract:Public auditing enables auditors to remotely verify data integrity for the outsourced data, which is an essential security issue and a promising solution for reliable cloud storage. However, in cloud storage systems, most existing public auditing schemes adopt a static auditing policy in the blockchain network, so that they are not able to adapt the dynamic environment efficiently, i.e., dynamic attacks, users joining and leaving, etc. Moreover, it is hard to improve the scalability with a static auditing policy, which may result in low performances and high security risks for the blockchain-based public auditing. In order to overcome the above limitations, we present a deep reinforcement learning-based method to improve the efficiency (i.e., transaction throughput and network latency) for the current blockchain-based public auditing solutions. Specifically, an blockchain-based security protocol is firstly proposed to guarantee that the integrity of outsourced data can be verified based on the dynamically auditing policy. Then, re-auditing time interval, the number of public auditors and the size of blocks are adjusted by the proposed deep reinforcement learning-based method to improve performance and security in a long term. Finally, security analysis indicates that the proposed work is able to resist the malicious entities and attacks derived from the Proof-of-Work consensus mechanism. Results of experiments that conducted in this paper demonstrate that the proposed scheme outperforms the baseline schemes on both transaction throughput and network latency.